Back to skill

Security audit

complex-bug-debugging-with-ai

Security checks for vulnerabilities and agentic risk

Overview

This is a disclosed debugging workflow skill with some over-prescriptive model and tool-access guidance, but no hidden execution, exfiltration, or deceptive behavior was found.

Before installing, treat the model requirement as advice rather than a mandate. When using the skill on real systems, approve production changes explicitly, limit DB/SSH/K8S/log access to the incident scope, and review any generated BUGxx.md for sensitive data before keeping or sharing it.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (1)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly requires a specific model ('Opus 4.7') and instructs the AI to stop or redirect the user if that model is not in use. This can improperly constrain user choice, override platform routing/policy expectations, and create coercive behavior not strictly necessary for the debugging workflow. In context, the skill is not trying to exfiltrate data or execute code, but it does embed hard model-selection pressure as a gate to service, which makes the issue real though less severe than an actively malicious prompt.

Content

No source excerpt is available for this finding.